What Drives Merger Waves?

1
What Drives Merger Waves?

• Jarrad Harford

2
Motivation

• The question of what drives waves of merger
  activity is an old one.
• Recently, authors such as Shleifer and Vishny and
  Rhodes-Kropf and Vishwanathan have developed
  behavioral models explaining the observed
  relation between stock market valuations and
  merger activity.
• This has refocused the debate as one between
  neoclassical and behavioral explanations for
  waves

3
Research Question

• Is a clustering of mergers at the aggregate level
  due to a combination of industry shocks for which
  mergers facilitate change to the new environment,
  or is it due to market timing?

4
Hypotheses

• Neoclassical hypothesis
• Based on the work of Coase (1937), Gort (1969),
  Maksimovic and Phillips (2001), Jovanovic and
  Rousseau (2001,2002)
• Economic disturbance leads to industry
  reorganization
• Modify to include a role for capital liquidity
• Eisfeldt and Rampini (2003)
• Shleifer and Vishny (1992)
• Once a technological, regulatory or economic
  shock occurs, the collective reaction of firms
  inside and outside the industry reallocates
  industry assets through merger and partial-firm
  acquisitions.

5
Hypotheses

• Behavioral Hypothesis
• Shleifer and Vishny (2003)
• Bull markets lead groups of bidders with
  overvalued stock to use it to buy real assets of
  undervalued targets.
• Model relies on dispersion in valuations and time
  horizons.
• Rhodes-Kropf and Viswanathan (2003)
• Rational targets accept bids from overvalued
  bidders because they overestimate synergies
  during market peaks.
• Both models rely on bidder managers timing market
  valuations.

6
Predictions

• Neoclassical Hypothesis
• Observable shocks will precede waves
• MOP will be either stock or cash
• Both mergers and partial-firm cash transactions
• Credit constraints will be low and/or asset
  values will be high

7
Predictions

• Behavioral Hypothesis
• Merger waves occur following periods of
  abnormally high stock returns or M/B ratios,
  especially when dispersion in those metrics is
  high
• Post-wave returns should be poor
• Post-merger operating performance should be
  particularly poor in waves
• MOP should be overwhelmingly stock
• Partial-firm transactions for cash should not be
  common

8
Sample

• I end-up with
• 35 waves from 28 industries (7 of which have 2
  waves, one in the 1980s and one in the 1990s)
• Average 24-month non-wave period has 7.8 bids
• Average 24-month wave period has 34.3 bids
• How I got there

9
Sample

• Start with all merger transactions (gt50mil) on
  SDC from 1981 to 2000
• Measure highest 24-month concentration in 1980s
  and 1990s for each industry (Fama-French 48
  industries)
• For each industry, compare that concentration to
  the empirical distribution of concentrations from
  1000 simulations specific to that industry
• If the actual concentration was greater than the
  95th percentile concentration in the
  distribution, categorize that period as a wave

10
Industry Characteristics and Waves

• Operating Environment
• Shocks to Profitability (cash flow/sales), Asset
  turnover (Sales/TA), Sales growth, ROA, Employee
  Growth,
• RD, Capital Expenditures
• Market valuations
• 1- and 3-year returns and s of those returns
• Market-to-book ratios
• Measure all variables in year prior to start of
  an industry merger wave

11
Univariate Results

• Summary of Table 2
• Shocks to profitability, asset turnover, sales
  growth, ROA, employee growth, RD, Capital
  Expenditures are all abnormally high
• M/B, change in M/B and s(M/B) are all abnormally
  high
• 3-year return is abnormally high, but 1-year
  return, s(3-year return), s(1-year return) are
  all not

12
Univariate Results

• Deregulation and Capital Liquidity
• Index of deregulatory events
• 4-quarter moving average of the Commercial
  Industrial rate spread
• Federal reserves SLO survey
• Highly correlated with the question of tightening
  or easing credit

13
Figure 1 Capital Liquidity, Industry Merger
Waves and Aggregate Merger Activity
14
Predicting an Industry Merger Wave

• Logit models
• All 48 industries for the 20-year sample period
• Problem operating shocks are too highly
  correlated
• Solution 1st principal component from 7 shock
  variables
• Include on its own and with an interaction for
  capital liquidity
• Low capital liquidity high CI rate spread or
  low M/B

15
Table 4 Predicting Merger Waves
16
Summary of Logit Models

• On their own, behavioral variables have some
  predictive power for merger waves.
• However,
• Commonly cited association between stock price
  performance and merger waves disappears after
  controlling for capital liquidity
• Once nested within a neoclassical model of merger
  waves, the behavioral variables add little
  explanatory power

17
Partial-firm (divisional) Acquisitions

• Neoclassical hypothesis predicts that
  transactions will occur both at the firm and
  divisional level and that the MOP will be both
  cash and stock
• Mulherin and Boone (2000), Maksimovic and
  Phillips (2001), Schlingemann, Stulz and Walkling
  (2002),
• Behavioral hypothesis cannot explain divisional
  transactions for cash, especially when executed
  by stock-swap bidders

18
(No Transcript)
19
 
Table 5 The Relation between Merger Activity and
Partial-Firm Acquisitions
20
Table 5, Panel B
21
Summarizing Partial-firm acq. results

• Total firm-level and partial-firm level activity
  are highly correlated
• Stock-swap only merger activity is correlated
  with cash partial-firm level activity
• Being a bidder strongly predicts being a buyer in
  a partial-firm transaction
• Even being a stock bidder strongly predicts being
  a cash buyer in a partial-firm transaction

22
Long-run returns

• Behavioral hypothesis predicts poor long-run
  returns following merger waves.
• Extant evidence is mixed (e.g. Loughran and Vijh,
  1997 Mitchell and Stafford, 2000)
• Use Fama (1998)s calendar time approach
• Create monthly portfolios of treatment firms
• Regress vector of monthly returns on FF 3-factor
  realizations

23
Table 6 Long-run Returns
24
Table 6, Panel B
25
Summary of Long-run Return Results

• There is little evidence of abnormally low
  returns following merger waves
• There is some evidence of poor post-merger
  returns for stock bidders in merger waves. This
  evidence is not robust to equal-weighted returns.

26
Operating Performance

• Inherently noisy tests during merger waves
• Idea is to compare with-merger performance to
  performance that would have occurred without the
  merger
• Pre-merger performance and industry performance
  are probably both noisy proxies for performance
  without the merger
• Consequently, neoclassical hypothesis makes no
  prediction about the observed post-merger
  performance in a wave
• Under the null of the behavioral hypothesis,
  post-merger performance in waves should be
  particularly poor

27
Table 7 Operating Performance Changes following
Mergers
28
Summary of Operating Performance Tests

• No evidence that post-merger operating
  performance is worse for mergers during waves
• Only evidence of a difference for wave vs.
  non-wave mergers is better post-merger asset
  turnover and M/B

29
Summary of Predictions and Findings
30
Conclusions

• Comparing the behavioral and neoclassical (with
  capital liquidity) hypotheses, the results
  consistently support the neoclassical hypothesis
• The role of capital liquidity causes industry
  merger waves to cluster in time even if industry
  shocks do not
• The relation between asset values and merger
  activity reflects capital liquidity rather than
  misvaluation

31
Conclusions

• While there is undoubtedly some merger activity
  driven by managers timing the market, such
  mergers do not cause merger waves.
• Aggregate merger waves are caused by the
  clustering of shock-driven industry merger waves,
  not by attempts to time the market.